A SUBTLE INFRARED EXCESS ASSOCIATED WITH A YOUNG WHITE DWARFIN THE EDINBURGH-CAPE BLUE OBJECT SURVEY

E. Dennihy1, John H. Debes2, B. H. Dunlap1, P. Dufour3, Johanna K. Teske4,5, and J. C. Clemens11 Physics and Astronomy Department, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA;

edennihy@unc.edu, bhdunlap@unc.edu, clemens@unc.edu2 Space Telescope Science Institute, Baltimore, MD 21218, USA; debes@stsci.edu

3 Institiut de Recherche sur les Exoplanétes (iREx), Département de Physique, Université de Montréal,Montreal, Québec H3C 3J17, Canada; dufourpa@astro.umontreal.ca

4 Carnegie DTM, 5241 Broad Branch Road, NV, Washington, DC 20015, USA; jteske@carnegiescience.eduReceived 2015 October 14; revised 2016 August 12; accepted 2016 August 13; published 2016 October 25

ABSTRACT

We report the discovery of a subtle infrared excess associated with the young white dwarf EC 05365–4749 at 3.35and 4.6 μm. Follow-up spectroscopic observations are consistent with a hydrogen atmosphere white dwarf ofeffective temperature 22,800 K and log [g (cm s−2)]=8.19. High-resolution spectroscopy reveals atmosphericmetal pollution with logarithmic abundances of [Mg/H]=−5.36 and [Ca/H]=−5.75, confirming the whitedwarf is actively accreting from a metal-rich source with an intriguing abundance pattern. We find that the infraredexcess is well modeled by a flat, opaque debris disk, though disk parameters are not well constrained by the smallnumber of infrared excess points. We further demonstrate that relaxing the assumption of a circular dusty debrisdisk to include elliptical disks expands the widths of acceptable disks, adding an alternative interpretation to thesubtle infrared excesses commonly observed around young white dwarfs.

Key words: circumstellar matter – planetary systems – white dwarfs

1. INTRODUCTION

The discovery of infrared excesses around white dwarf starsand their interpretation as debris disks have opened an excitingnew avenue for planetary research. Atmospheric metals inotherwise pristine hydrogen atmosphere white dwarf stars arein most cases now understood to be delivered to the surfacefrom dust disks formed from tidally disrupted planetary bodies(Debes & Sigurdsson 2002; Jura 2003; Veras et al. 2014). Theextreme stargazing orbits required for the tidal disruption ofrocky exo-planetary material provide an interesting frameworkfor studies on post-main sequence planetary evolution throughplanetary and binary interactions (Debes et al. 2012; Bonsor &Veras 2015). Detailed abundance studies of white dwarfatmospheres have shown signatures of carbonaceous asteroids,planetary lithospheres, and even water (Zuckerman et al. 2011;Farihi et al. 2013; Jura et al. 2015). A subsample of dusty disk-hosting white dwarfs with gaseous disk components have alsoshown dynamical evolution on timescales of years (Wilsonet al. 2015). These comprehensive studies offer insight intoexoplanetary chemical composition and dynamics not acces-sible to other exoplanet studies.

The systematic search for these ideal exoplanetary labora-tories saw early progress with the targeted infrared photometryof known metal-polluted white dwarf stars (Kilic et al. 2006).The Spitzer Space Telescope was invaluable for confirmingmany disks and constraining their radii to within the tidaldisruption radius of their host white dwarf stars (Jura et al.2007; Farihi et al. 2008). Later, all sky near-infrared surveyssuch as the Two Micron All Sky Survey (2MASS; Skrutskiet al. 2006) and the Wide-field Infrared Sky Survey (WISE;Wright et al. 2010) provided a basis for rapidly surveyingknown white dwarfs for infrared excesses (Debes et al. 2011;Hoard et al. 2013). To date, these techniques have increased the

sample of debris disk hosting white dwarfs to nearly 40confirmed systems (Rocchetto et al. 2015). The continuedgrowth of the sample is critical for exploring the full range ofphenomena that evolved planetary systems exhibit.In this paper we present the discovery of an infrared excess

at the DA white dwarf EC 05365–4749 (hereafter EC 05365),as well as the detection of atmospheric metal pollution in itsoptical spectra. Originally identified in the first two zones of theEdinburgh-Cape Blue Object Survey (hereafter EC survey;Kilkenny et al. 1997; O’Donoghue et al. 2013), EC 05365 isidentified in several all-sky photometric surveys at 5h37m53 5,−47°58′05 3 in FK5 coordinates. In Section 2 we discuss thephotometry and spectroscopic follow-up collected to constructan informative spectral energy distribution (SED) and modelthe atmospheric pollution. Section 3 begins by establishing theinfrared excess, followed by a range of modeling effortsincluding relaxing the assumption of a circular debris disk toinclude elliptical debris disks. Section 3 closes with an analysesof the atmospheric pollution and accretion rates. Section 4places EC 05365 in the context of other dusty disk-hostingwhite dwarfs and presents recommendations for follow upobservations.

2. TARGET SELECTION AND OBSERVATIONS

Following the prescriptions laid out by Debes et al. (2011)for the WISE Infrared Excess around Degenerates survey, wecross-correlated the DA atmosphere white dwarfs identified inZones 1 and 2 of the EC survey with the ALLWISEphotometric catalog and 2MASS All Sky Data Release PointSource Catalog (2MASS PSC). Many, but not all, of the whitedwarfs identified in the first two published zones of the ECsurvey have been incorporated into the McCook and Sion whitedwarf catalog (McCook & Sion 1999) and were included in theWIRED catalog by Hoard et al. (2013). This search was carriedout with the missing targets in mind. We will extend our search

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 doi:10.3847/0004-637X/831/1/31© 2016. The American Astronomical Society. All rights reserved.

5 Carnegie Origins Fellow, jointly appointed by Carnegie DTM & CarnegieObservatories

1

to include the remaining EC survey zones as they arepublished. Of the approximately 110 targets surveyed in thispaper, two were identified as strong excess candidates:EC 05365, the focus of this paper, and WD 1150–153, apreviously known dusty white dwarf (Kilic & Redfield 2007).These numbers are roughly consistent with recent estimates ofthe frequency of dust disks around white dwarfs (Debes et al.2011; Barber et al. 2014; Rocchetto et al. 2015).

2.1. Photometry

We collected published photometric measurements ofEC 05365 from the 2MASS PSC (Cutri et al. 2003), theALLWISE photometric catalog (Cutri et al. 2013), the AAVSOPhotometric All Sky Survey (APASS) (Henden et al. 2009), theGalaxy Evolution Explorer (GALEX; Martin et al. 2005) DataRelease 6, and the VISTA VHS 3rd data release (McMahonet al. 2013), enabling the construction of an SED spanningultraviolet to near-infrared wavelengths. We also collectednear-infrared images for astrometric information from theVISTA Science Archive, described in Irwin et al. (2004),Hambly et al. (2008), and Cross et al. (2012).

As discussed in Debes et al. (2011), the large point-spreadfunction of the WISE telescope often leads to contamination orsource confusion. High spatial resolution follow-up images arenecessary for identifying nearby contaminants. The WISEphotometry system implements profile fit photometry andsource deblending routines that are robust against contamina-tion of resolved sources separated by more than 1.3×FWHM(approx. 7.8 arcsec for W1) of the point-spread function of theband of interest, and is capable of flagging sources con-taminated by a nearby unresolved source by virtue of thegoodness of fit of the best-fit point-spread profile. The use ofthese deblending routines in the photometric measurement of asource is captured in a set of photometric flags; “nb” for thepassive deblending routine of resolved sources separated bymore than 1.3×FWHM of the point-spread function of theband, and “na” for the active deblending of a pair of unresolvedsources. For the photometry taken at the source positionassociated with EC05365, the source was not flagged for activeprofile deblending, indicating that the reported photometry isnot corrected for contamination for a nearby source within7.8 arcsec.

Figure 1 displays the VISTA VHS Ks band image whichallowed us to identify two potential nearby contaminants,Source A and Source B. The red circle indicates the limit ofsource resolution implemented by the ALLWISE photometrypipeline prior to attempting blended profile fit photometry. TheVISTA J and Ks band measurements and approximateseparation from the ALLWISE detection photocenter for eachsource are given in Table 1. Source B is just at the limit ofresolution. Given the relatively faint Ks magnitude and J−Ks

color, we suspect it is below the detection limit of the WISEbands and not a source of contamination for the ALLWISEdetection. Source A, however, is comparable in Ks brightnessand has a J−Ks color, indicating it could have a measurableflux in the WISE 1 and 2 bands. Source A is also extended andhas a high probability of being extragalactic. Considering onlythe information contained in the ALLWISE catalog, we haveno way of determining whether the ALLWISE measurementrepresents a blend of the target and Source A, or if it is insteadconsistent with a single source at either the position of thetarget or Source A. However, given that we know the positions

of the target and Source A very well (compared to the coursesampling of the WISE pixel scale) from the VISTA catalog, wecan perform a set of estimations to determine the most likelycase. We discuss this potential for contamination in detail in thecontext of our SED in Section 3.2.

2.2. Spectroscopic Follow-up

Photometry alone is sufficient to identify the infrared excess,but spectroscopic follow-up is necessary for identifying andcharacterizing the atmospheric pollutants that indirectly probethe remnant planetary system. We first observed EC 05365with the Goodman spectrograph (Clemens et al. 2004) on theSOAR telescope configured to search for atmospheric metalpollution on 2015 April 04. The 0 46 long slit and 1800 l/mmgrating were chosen to maximize resolution (≈0.66Å perresolution element) while allowing the spectral range to probefor the Ca II H &K transitions at 3934 and 3967Å, and theMg II doublet at 4481Å. We achieved a signal-to-noise ratio(S/N) per resolution element of 27 near 4000Å in 3240 s ofcombined exposure time. The data revealed an absorptionfeature at the approximate location of the Mg II doublet with anequivalent width (EW) of 75±16 mÅ, though without apublished radial velocity or suitable comparison line we couldnot rule out the possibility that the feature was interstellar.To confirm the detection of Mg and search for additional

metal species, we observed EC 05365 with the MIKEspectrograph on the Magellan telescope on 2015 April 30.

Figure 1. Infrared Ks image taken from the VISTA VHS Survey DR3.EC 05365 is identified at 5 37 53. 5h m s , −47°58′05 3 in FK5 coordinates alongwith a nearby sources “A” and “B.” A 7.8 arcsec circle is shown centered onthe red cross which identifies the ALLWISE detection photocenter.

Table 1VISTA VHS Photometry

Object J (mag) sJ Ks (mag) sKs Sep (″)

EC 05365 16.099 0.010 16.114 0.038 1.12Source A 17.869 0.041 16.615 0.057 4.46Source B 17.796 0.036 17.469 0.128 7.73

2

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.

We used the 0 5×5″ slit which translated to a resolution of0.083Å per resolution element and a S/N per resolutionelement of 25 near 4000Å in the combined 3×1200 sexposures. No standard star was observed as the goal wasstrictly to confirm atmospheric pollution measured relative tothe local continuum. The data were extracted and flatfieldedusing the MIKE reduction pipeline, with methodologydescribed in Kelson et al. (2000) and Kelson (2003). Weconfirmed the detection of Mg II in the Goodman data with anEW of 83±6 mÅ and discovered a Ca K absorption featurewith an EW of 47±11 mÅ. Both features were found to be atthe photospheric velocity as calibrated against the non-LTEcore of the hydrogen alpha absorption line. After applyingheliocentric corrections we found the photospheric velocity(combined gravitational redshift and radial velocity) to be40±7 km s−1.

Finally, to better constrain the atmospheric parameters weobserved EC 05365 with the Goodman spectrograph on 2015May 21 with a low-resolution setup. The spectrograph wasconfigured with the 930 l/mm grating and the wide 3 0 longslit to minimize the effect of variable seeing conditions, whichvaried between 1 4 and 1 6. The seeing-limited resolution was≈4.2Å per resolution element with a wavelength range of 3700to 5200 Å and we achieved a S/N of 110 per resolutionelement near 4200Å in the combined 7×180 s exposures. Wereduced and extracted the spectra with standard IRAFtechniques and employed user-developed techniques forwavelength and continuum calibration. The spectroscopicstandard LTT 2415 was observed to provide continuumcalibration but, as discussed in Section 3.1, cloudy conditionsrendered the standard unreliable for flux calibration.

3. ANALYSIS AND DISCUSSION

3.1. Atmospheric Parameters

Though the atmospheric parameters of white dwarfs can beconstrained with photometric data (Bergeron et al. 1995a), thesurface temperature implied by our photometry is poorlyconstrained (21,000–26,000 K), which is often the case forstars this hot. We instead rely on the spectral line fittingtechnique initially developed by Bergeron et al. (1992) andrefined by Bergeron et al. (1995b) and Liebert et al. (2005).The method generates a grid of model spectra over a range ofsurface gravities ( glog ; g measured in cm s−2) and effectivesurface temperatures (Teff degrees K). A Levenberg–Marquardtchi-squared minimization is then used to determine whichvalues of glog and Teff best reproduce the Balmer profilesrelative to their local continuum. By using only the shape anddepth of the Balmer profiles, this method is expected to berobust against overall flux calibration errors.

We first obtained atmospheric parameters using the low-resolution Goodman data and the corresponding spectro-photometric standard data for flux calibration. The best-fit

glog and Teff were 8.03±0.05 and 24 050±150 K with theerror bars representing the statistical uncertainties of the fittingprocess. As expected, the individual Balmer profiles werereproduced well by the model atmosphere but the continuumshape between lines was poorly fitted. Given the cloudyconditions of the night, we suspected the data used for fluxcalibration to be unreliable.

To asses the reliability of the standard flux calibration, wegenerated a comparison set of spectra which were flux

calibrated with an independent set of white dwarf modelspectra, kindly provided by D. Koester (see Koester 2010 fordetails on model input physics and methods). We chose modelsspanning the photometrically constrained temperature rangeand only used continuum points for calibration to avoid biasingthe fits to the Balmer profiles. When fitting the model fluxcalibrated spectra of EC 05365 we found the best fitting glogand Teff to be 8.20±0.05 and 22 800±150 K independent ofthe white dwarf model used for flux calibration. The agreementof the model flux calibrated solutions further supports oursuspicion of the standard flux calibrated data. We thereforeadopt = -

+T 22 800eff 1501400 K and = -

+glog 8.19 0.200.05 for the

remainder of the analysis, incorporating the solutions fromour standard flux calibrated data into our reported uncertainty.Figure 2 shows the Balmer series fits to the model calibratedGoodman data. Using the evolutionary sequences of Fontaineet al. (2001), the surface gravity and temperature are consistentwith a white dwarf mass and cooling age of -

+0.74 .12.03

Mof -

+63 367 Myr.

3.2. Establishing the Infrared Excess

First, we converted all measured photometry to units of fluxdensity using published zero points (Holberg & Bergeron 2006;Skrutski et al. 2006; Wright et al. 2010). We scaled the whitedwarf model to the observed photometry by calculating amedian scale factor for all photometry points without clearexcess (�2.15 μm). The scaling factor is consistent with aphotometric distance of -

+87 313 parsecs. The resulting scaled

white dwarf model is plotted with the observed photometry inFigure 3.

Figure 2. Normalized hydrogen balmer lines of spectroscopic data taken withSOAR with best-fitted glog and Teff model overplotted in red.

3

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.

The departure of the W1 and W2 photometry from the whitedwarf model of EC 05365 is evidence for an external infraredbright source. But before we can attribute it to a sourcecoincident with EC 05365 we must consider the contribution totheW1 andW2 measurement of the nearby extragalactic SourceA. Since there were no photometric flags indicating profile,deblending was not performed for this detection, and anycontributions from Source A are included in the catalogmeasurement.

We are unable to model the expected contribution of thegalaxy to W1 and W2 bands as it is only reported in the twoVISTA bands and we cannot constrain both the spectral typeand redshift. Instead we used the galaxy SED atlas of (Brownet al. 2014) to determine whether any nearby galaxy SEDscould match the observed J−Kscolors and reproduce theentire W1 and W2 excess. The atlas constitutes a large set ofobserved SEDs of nearby (z<0.05) galaxies spanning a widerange of morphological types. Few of the observed catalogSEDs are able to reproduce the observed Source A slope from Jto Ks and none of them are likewise consistent with the W1 andW2 excess. We are confident this rules out the possibility of anearby galaxy accounting for the excess, but cannot rule ahigher redshift source.

A by-eye extrapolation of the flux from Source A in Figure 3might suggest a very strong contribution to the W1 and W2measurements, but the SED does not consider all of theinformation available, namely the relative separations of theALLWISE photocenter from the measured positions ofEC 05365 and Source A in the VISTA VHS catalog. Basedon the additional astrometric evidence, we now argue that evenif there is contamination from Source A in the W1 and W2measurements, there is still a significant infrared excessassociated with EC 05365.

In Figure 4 we show the VISTA Ks and the WISE W2 bandimage, overplotted with isoflux contours of the W2 image. It isimmediately apparent that the contours of the WISE detectionare more consistent with a source detection at the positionof the target EC05365, with a minor asymmetry and offset inthe direction of Source A suggesting some measureable

contamination. As a first-order estimate of the level ofcontamination, we can use the relative separations of the targetand Source A from the WISE photocenter as a rough proxy oftheir relative flux. Assuming the positions of the target andSource A as reported by the VISTA Science Archive andconsidering the source position of the ALLWISE detection andits associated error ellipse, we find the ratio of the flux ofthe target to the flux of Source A within the WISE bands to be4.0/1.0±0.6. If we subtract this level of contribution ofSource A from the ALLWISE flux measurements (approx.20%) we still find an excess above the white dwarf model in theW1 and W2 bands of 13σ and 12σ, assuming the reportedphotometric errors. Given the significance of the excess, we donot expect any systematic error introduced by assuming thereported photometric errors to invalidate the excess.As a more informative approach, we followed the technique

of “forced” photometry as demonstrated by Lang et al. (2016).In essence, the technique relies on forward modeling of theWISE catalog images taking the positions of sources, the point-spread function of the band of interest, and a per pixel noisemodel as knowns, and solving for the most likely combinationof source fluxes that can reproduce the observed data (Langet al. 2016).For our images we use the unWISE coadds produced by

Lang (2014), which are a publicly available set of WISE coaddssampled at the nominal resolution of the detector and optimallycombined for the purpose of forced photometry. As our point-spread functions for each WISE band, we use a combination ofGaussian functions as derived in Lang et al. (2016) with weightand standard deviation given in their Table 1. As an estimate ofthe local noise we choose three neighboring regions that aresource free in the VISTA J and Ks images, and in the unWISEcoadded images. Since we are only concerned with the relativeflux of two sources, we proceed by generating a grid of fluxesfor each source. At each point on the grid we convolve thefluxes at their source position with the point-spread function ofthe band of interest, sampled at the resolution of the image, andsubtract the result from the unWISE image. We then comparethe resulting subtracted image with the local noise estimatesusing the method of least squares to find the most probablecombination of fluxes. We perform this exercise separately onthe W1 and W2 images.As expected, the result of the forced photometry measure-

ment varies between W1 and W2. For the W1 band, a target toSource A flux ratio of 2.0/1.0 best reproduced the image.Whereas for the W2 band, the best fitting target to Source A

Figure 3. EC 05365 observed photometry in blue overplotted with EC 05365white dwarf model shown in light gray. The VISTA photometry for Source Aand the spatially coincident ALLWISE photometry are plotted separately in redand green. Note that the errors presented are catalog uncertainties and likelyunderestimate the absolute (including systematic) photometric uncertainty.

Figure 4. Left: VISTA Ks image with W1 isoflux contours overlaid. Right:WISE W1 image with isoflux contours. The images are 18″×20″ to match thescale of Figure 1. Both the minor asymmetry of the contours along the linebetween EC 05365 and Source A and the offset in the ALLWISE photocenterfrom the position of EC 05635 suggest potential for source confusion.

4

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.

flux ratio was 3.39/1.0. Assuming that Source A contributesthis fraction of flux to the ALLWISE detection, if we againsubtract the contribution from Source A to the W1 and W2measurements we find the remaining flux above the whitedwarf model to represent 8σ and 11σ excesses respectively.

3.3. Modeling the Infrared Excess

The complete photometry set is collected in Table 2. Notethat the errors presented are catalog uncertainties and likelyunderestimate the absolute (including systematic) photometricuncertainty. Given the uncertainty in our WISE contaminationestimates and the lack of near-infrared data for modeling, we donot expect inflated error bars to have much effect on ourconclusions. We proceed by exploring different astrophysicalmodels that can reproduce the infrared excess associated withEC 05365, in each case considering a minimally (using theresults from astrometry alone) and maximally (using the resultsfrom the forced photometry) contaminated flux measurementfor the W1 and W2 bands. For the remainder of the section wewill refer to these results as Case 1 and Case 2 levels ofcontamination.

The simplest model we explored was a spherical blackbodysource taken to be at the photometric distance of the whitedwarf, e.g., an unresolved low-mass companion. A minimumchi-square fitting method was used to determine a best-fittingtemperature and radius for the Case1 and Case 2 contaminationlevels of 1040 K, 0.15 R and 800 K, 0.23 R respectively. Thetemperatures are consistent with the T dwarf sub-stellar classbut the radii are implausibly large. In addition, the J−Ks colorprovided by both 2MASS ( J−Ks<0.556) and VISTA( - = -J K 0.28s ) rule out a companion of this spectral class(Chabrier et al. 2000).

By dividing the bolometric flux of these single-temperatureblackbody solutions with the bolometric flux of the modelwhite dwarf, we can also get an estimate for the fractionalinfrared luminosity t = L LIR of the system. Such a metrichas been demonstrated by Farihi (2011, p. 117) and Rocchettoet al. (2015) to point toward aggregate properties of the discs ofdusty white dwarfs. The different levels of contamination resultin a range of τ=0.066%–0.080%, adding EC 05365 to thegrowing sample of subtle infrared excesses around young whitedwarfs.

We also modeled the infrared excess with a flat, opticallythick circular disk using the formalism developed by Jura(2003). The model invokes a radially dependent temperaturedistribution and can be used to constrain the inclination, inner,and outer radius of the disk. There is however an unavoidabledegeneracy between inclination and the width of the disk,which is amplified in the absence of longer wavelength data.Having only two mid-infrared data points with clear excess, wefixed the inclination and explored only the inner and outer

radius of the disk. Considering the degeneracy between widthand inclination, we fitted the models under assumptions of bothhigh (i= 80°) and low (i= 45°) inclination, and present both asviable solutions.We used a minimum chi-squared fitting procedure to

determine the best fit inner radius and disk width. Weparameterized the inner radius in terms of its temperature andexplored a range of inner temperatures from 500 to 2500 K anda range of disk widths from 1 to 20 WD radii. The results fromconsidering both the Case 1 and Case 2 contaminated W1 andW2 measurements are summarized in Table 3. We overplot thebest-fitted combined SEDs of the white dwarf and the dustdisks on the observed photometry for Case 1 and Case 2contamination in Figure 5.The range of parameters that can fit the data is large, but

qualitatively the subtle excess forces the circular disk models totwo classes of solutions: a wide disk observed at highinclination or a narrow disk observed at low inclination. Withno prior expectation for the inclination, there is no preferencefor either class of solutions. The subtle excess associated withEC 05365 is so far typical of dusty white dwarfs with coolingages below 200Myr, as recently explored by Rocchetto et al.(2015). The authors demonstrate that the assumption of diskswith dust populating the entire available region between thesublimation radius, where dust particles are destroyed by thestellar radiation, and the tidal disruption radius, where the dustis produced, is inconsistent with an assumption of randomlyinclined disks, given the high inclinations needed to model thesubtle infrared excesses commonly observed for younger dustywhite dwarfs. The tendency toward narrow disks (which do notspan the available space between the sublimation radius andtidal disruption radius) is less apparent for older dusty whitedwarfs with cooling ages between 200 and 700Myr, presentingan interesting delineation between the two samples.As a case study, we explored whether the necessity of

narrow disk width for low inclinations could be relieved forEC 05365 by modeling the infrared excess as a set ofconfocally nested elliptical rings rather than concentric circularrings. This was motivated by the fact that the formationscenario of dusty debris disks necessitates highly eccentricorbits during the tidal disruption phase. The eccentric ringsformed by individual disruptions are expected to circularizeunder the effects of Poynting–Robertson drag on timescalesranging from tens to thousands of years for the relevant particlesize scales (Veras et al. 2015), but the effects of multipledisruption events and collisions on the eccentricity evolution ofdisks is largely unknown.We begin by adopting the same passive, flat, and opaque

assumptions about the dust particles as considered by Jura(2003). We consider the white dwarf to be at the center of ourradial coordinate system, and the inner and outer ellipses that

Table 2EC 05365 Collected Photometry

GALEX APASS 2MASS VISTA WISEa

FUV NUV B g V r J H Ks J H Ks W1b W2b

Flux (mJy) 7.51 4.78 2.66 2.44 2.24 1.87 0.49 0.44 L 0.556 L 0.239 0.141–0.171 0.147–0.152σ (mJy) 0.34 0.13 0.07 0.13 0.04 0.29 0.06 0.07 L 0.005 L 0.008 0.004 0.008

Notes. The errors presented are catalog uncertainties and likely underestimate the absolute (including systematic) photometric uncertainty.a The range of flux is bound by the difference between Case 1 and Case 2 contamination discussed in Section 3.2.b The original reported flux values for the potentially contaminated ALLWISE catalog source are 0.211 and 0.192 (mJy), respectively.

5

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.

bound the disk to be described by the parameters ain, aout, and ewhere ain and aout define the semimajor axes of the inner andouter ellipses, and e defines the eccentricity of the ellipseswhich we hold fixed for all nested ellipses. The ellipses areconfocally nested with the white dwarf at one focus. Anillustration of this geometry is shown in Figure 6. In this way,the coordinate r specifies the distance to a dust particle from thewhite dwarf which uniquely determines the particle temper-ature under the assumption of an optically thick, flat disk. Theinner/outer ellipses which bound the disk are then described bythe equation:

qq

=-

-r

a e

e

1

1 cos1in out

in out2

( )( )

( )

and our integral for the observed monochromatic flux as afunction of frequency becomes:

ò òq=np

q

qnF

i

Dd B T r rdr

cos2

r

r

2 0

2

ringin

out

( ( )) ( )( )

( )

where i is the inclination of the disk and D is the distance of thesystem.

For a qualitative comparison between circular and ellipticalmodels, we applied both models to the Case 2 contaminationexcess observed for EC 05365 with a fixed inner radius of thedisk corresponding to a temperature of 1100 K. We varied thedifference between the periastron of the inner/outer ellipsesand the inclination for both a circular (e=0.0) and highlyelliptical (e=0.95) disk model to produce the constant chi-

square boundaries shown in Figure 7. The 1-σ, 2-σ, and 3-σconfidence intervals correspond to the chi-squared contours of2.30, 6.18, and 11.8 above the minimum chi-squared value foreach grid, as appropriate for two parameters of interest. Thecircular and highly eccentric models have comparable mini-mum chi-squared values which are consistent with a reason-able fit.There is a stark difference between the acceptable parameter

space for EC 05365 for circular and elliptical disks illustratedby Figure 7: for a given inclination, the elliptical disk ismarkedly wider at periastron. This can be intuitively under-stood as close-in material is re-distributed to farther distancesalong the eccentric orbit, effectively shifting flux from the near-infrared out to longer wavelengths.This qualitative difference holds true independent of the

inner dust temperature assumed. Given the lack of long-wavelength data where the models begin to diverge, there is noquantitative (e.g., goodness-of-fit measurement) distinctionbetween either model, especially considering the addition ofthe free parameter e. The results are, however, encouraging forelliptical SEDs as an alternative explanation for the lack ofyounger disks with dust spanning the available orbital rangebetween the sublimation and tidal disruption radius. Thedelineation between the dust distributions of the younger andolder samples could instead be the product of a third parameter,eccentricity. We explore this possibility and the application ofelliptical dusty rings to the larger dusty white dwarf sample in aforthcoming paper.

3.4. Abundance Measurements and Ratios

We followed the procedures detailed in Dufour et al. (2010)to estimate the atmospheric abundance of Ca and Mg. Inshort, assuming the atmospheric parameters above, wegenerated model white dwarf atmospheres using the codesdeveloped by Dufour et al. (2007) and performed a chi-squaredminimization over a grid of atmospheric abundances for

Table 3Best-fitted Circular Dust Disk Parameters

Fixed i Tinner Touter Rinner Router(deg) (K) (K) ( R ) R )

Case 1 Contamination:45 1100 1030 34 3780 1150 860 32 47Case 2 Contamination:45 860 800 47 5280 920 650 43 68

Figure 5. Best-fitted circular disk models overplotted on measured photometryand white dwarf model. The minimally contaminated solutions are shown inred while the maximally contaminated solutions are in blue. The solid/dashedlines represent low/high-inclination solutions.

Figure 6. An illustration of geometric differences between disk models withfixed periastron width and increasing eccentricity. The white dwarf is shown asthe filled black circle and the blue, red, and light green dashed/solid linesrepresent inner/outer bounding curves for disks with eccentricities of 0.0, 0.5,and 0.95.

6

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.

each element detected. We find the atmospheric metal pollutionto be consistent with logarithmic abundances of [Mg/H]=−5.36±0.25 and [Ca/H]=−5.75±0.25. Figure 8 showsthe best fit profile to the Mg 4481Å doublet.

The observed abundances can be transformed into estimatesof the accretion rate with knowledge of the rates of diffusionfor each element. Using the atmospheric parameters derived inSection 3.1 and the updated diffusion calculations described inKoester (2009),6 we determined the diffusion timescales for Caand Mg to be ´ -2.75 10 3 and ´ -4.47 10 3 years respectively.The atmospheric calculations of Koester (2009) also include anestimate of the hydrogen mass necessary to translate theobserved photospheric number densities into mass densities forthe accretion rates of each species. Assuming the hydrogenmass to an optical depth of τ=2/3 is ´ -1.73 10 17

M , wefind the system to be accreting Ca at a rate of ´3.8 107 g s−1

and Mg at rate of ´2.6 107 g s−1. We did not include theeffects of radiative levitation on the observed abundance andinferred accretion rate as they are are less important for lighterelements such as Ca and Mg, and negligible for accretion ratesas high as observed (Chayer & Dupuis 2010).

The most exciting feature of the dusty metal-polluted whitedwarfs is the opportunity to translate the observed abundancesinto those of the accreted parent body, thereby probing rockyexoplanetary abundances not accessible to other directmethods. The observed abundances depend on the rates ofaccretion, diffusion, and critically the understanding of thecurrent accretion phase. As discussed in Koester (2009), therate of diffusion sets the relevant timescale for the accretionphase and it is reasonable to expect the atmosphericabundances have reached a steady state as long as the accretionhas been held constant for >5 tdiff . In the case of EC 05365, thediffusion timescales for both Ca and Mg are on the order ofdays, guaranteeing that accretion has proceeded for severaltimes the diffusion time. Therefore, we can be certain theatmospheric abundance has reached a steady state and, towithin the ratios of the diffusion timescales, the observed

atmospheric abundance ratios directly reflect that of theaccreted parent body. Including the ratio of the diffusiontimescales derived above we calculate the parent abundanceratio as [Mg/Ca]=+0.24±0.25.Because these two elements have strong transitions in the

optical, the [Mg/Ca] ratio has been studied for a large sampleof polluted white dwarfs (see the recent assessment of Jura &Xu 2013,7 in particular their Figure 1). Despite somedispersion, it is clear that the ensemble abundance ratio reflectsbulk Earth composition, but the abundance ratio for EC 05365represents a large departure from the mean of this sample.One possibility for this departure could be the accretion of

some refractory dominated material, such as the crust of alarger, differentiated body. Such a scenario is discussed by Juraet al. (2015), and the spread in abundance ratios of well-studied

Figure 7. Dashed, dotted, and solid lines show filled contours of 1-σ, 2-σ, and 3-σ confidence intervals of the periastron width and inclination of a dust disk with innerdust temperature 1100 K around EC 05365. The left-hand panel shows how the near-infrared flux of the SED constrains the dust to be in either narrow rings or viewedat high inclinations. The right-hand panel shows how elliptical disks might allow a disk at periastron to be wider over a larger range of inclinations.

Figure 8. Magellan/MIKE data overplotted with the atmospheric abundancemodel around Mg II 4481 Å. The red line is the best-fit model with alogarithmic abundance of [Mg/H]=−5.36.

6 See updated tables here: http://www1.astrophysik.uni-kiel.de/~koester/astrophysics/astrophysics.html.

7 Jura & Xu (2013) use the instantaneous accretion approximation (alsodescribed as the early phase approximation as it does not account fordifferences in elemental diffusion timescales) for the sample in their Figure 1.Under this assumption log[Mg/Ca]=+0.39 for EC 05365.

7

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.

polluted white dwarfs is further evidence for the post-nebularprocessing of extrasolar asteroids. The closest analog toEC 05365 is GD 362, which Xu et al. (2013) demonstratedmust have accreted from either multiple distinct sources, orsome material that underwent post-nebular processing. Adetailed abundance study will be necessary to confirm whetherEC 05365 is indeed accreting from such a unique source.

4. CONTEXT AND FOLLOW-UP

As the sample of dusty white dwarfs grows, it is important toconsider the discovery of each system in the context of thelarger dusty white dwarf sample. EC 05365 is among thehottest and therefore youngest by cooling age of the dustywhite dwarfs. The disk parameters are not well constrained butthe fractional infrared luminosity L LIR =0.07% is quite low(Rocchetto et al. 2015), adding to the sample of subtle infraredexcesses around young white dwarfs. We have demonstratedfor EC 05365 that the application of an elliptical ring model canhelp relieve the requirement that the subtle excess be modeledby either narrow or highly inclined circular dust rings, and willconsider the application of elliptical rings to the larger sampleof dusty white dwarfs in a forthcoming paper.

Though the abundance ratio of [Mg]/[Ca] suggests theaccreted material is not strictly bulk Earth composition, forcomparison with the larger sample of known accreting whitedwarfs we scale the accretion rate of Ca assuming bulk Earthabundances and determine a total metal accretion rate of

= ´M 2.37 10Z9˙ g s−1, consistent with similarly young sys-

tems (Farihi et al. 2012).Detailed abundance studies of highly polluted dusty white

dwarfs are the most sensitive probes of rocky exoplanetarysystems. The combination of optical spectroscopy with largeground-based telescopes and ultraviolet spectroscopy with theCosmic Origins Spectrograph (COS) has proven effective atprobing the basic atomic abundances of silicate dust thatpresumably participated in early terrestrial planet formationaround the progenitors of these dusty white dwarfs. Morebroadly, hot DA white dwarfs with strong mid-infraredexcesses are particularly valuable in validating the connectionbetween dust abundance, dust accretion, and our understandingof the white dwarf photospheric diffusion process. Figure 9presents a collection of dusty white dwarfs arranged by bothGALEX FUV magnitude and WISE W2 magnitude. The roughmagnitude limits for spectroscopy with COS in the far-ultraviolet or the Mid InfraRed Instrument (MIRI) in themid-infrared identify 18 white dwarfs that can have detailedphotospheric abundances from the UV, and will also beamenable for detailed mineralogy of the dust residing in theirdisks via the JWST. These two measurements provideindependent confirmation of the abundance of the dust in thewhite dwarf photosphere, and that contained in the disk, suchas has previously been done only for G29-38 (Xu et al. 2014).We identify this sample, which now includes EC 05365, ascritical for future space-based follow-up studies.

We would like thank the anonymous referee for the detailedreports which greatly improved this manuscript. We would alsolike to thank Detlev Koester for graciously providing his whitedwarf atmospheric models. E.D. and J.C.C. acknowledge thesupport of the National Science Foundation, under award AST-1413001. This work is based on data obtained from (1) theWide-Field Infrared Survey Explorer, which is a joint project

of the University of California, Los Angeles, and the JetPropulsion Laboratory (JPL), California Institute of Technol-ogy (Caltech), funded by the National Aeronautics and SpaceAdministration (NASA); (2) the Two Micron All Sky Survey, ajoint project of the University of Massachusetts and theInfrared Processing and Analysis Center (IPAC)/Caltech,funded by NASA and the National Science Foundation(NSF); (3) the 6.5 m Magellan Telescopes located at LasCampanas Observatory, Chile; (4) the Southern AstrophysicalResearch (SOAR) telescope, which is a joint project of theMinistério da Ciência, Tecnologia, e Inovação (MCTI) daRepública Federativa do Brasil, the U.S. National OpticalAstronomy Observatory (NOAO), the University of NorthCarolina at Chapel Hill (UNC), and Michigan State University(MSU); (5) the VizieR catalog access tool, CDS, Strasbourg,France; (6) the NASA/IPAC Infrared Science Archive, whichis operated by JPL, Caltech, under a contract with NASA; (7)the NASA Astrophysics Data System; (8) observations madewith the ESO Telescopes at the La Silla or ParanalObservatories under programme ID(s) 179.A-2010Facilities: WISE, CTIO:2MASS, AAVSO (APASS),

Magellan:Clay (MIKE), SOAR (GOODMAN).

REFERENCES

Barber, S., Kilic, M., Brown, W., & Gianninas, A. 2014, ApJ, 786, 77BBergeron, P., Saffer, R., & Liebert, J. 1992, ApJ, 394, 228BBergeron, P., Wesemael, F., & Beauchamp, A. 1995, PASP, 107, 1047BBergeron, P., Wesemael, F., Lamontagne, R., et al. 1995, ApJ, 449, 258Bonsor, A., & Veras, D. 2015, MNRAS, 454, 53BBrown, M., Moustakas, J., Smith, J., et al. 2014, ApJS, 212, 18BChabrier, G., Baraffe, I., Allard, F., & Hauschildt, P. 2000, ApJ, 542, 464CChayer, P., & Dupuis, J. 2010, in AIPC 1274, 17th European White Dwarf

Workshop, ed. K. Werner & T. Rauch (Tubigen: AIP), 394Clemens, J., Crain, A., & Anderson, R. 2004, Proc. SPIE, 5492, 331C

Figure 9. A plot of all currently known dusty WDs with approximate COS andMIRI magnitude limits. Blue squares are WDs already observed with COSwhile green squares are WDs that have estimated far-ultraviolet magnitudesfrom their known Teff . The red diamond is EC05365. We also over plot roughmagnitude limits in purple lines for objects that can have moderate- to high-resolution ultraviolet spectroscopy (S/N>20 in 2 hr with COS G130M/1291) and moderate-resolution mid-infrared spectroscopy with the JWST (S/N>10 in 10 ks with MIRI/MRS).

8

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.

Cross, N. J. G., Collins, R. S., Mann, R. G., et al. 2012, A&A, 548A, 119Cutri, R., Skrutskie, M., van Dyk, S., et al. 2003, yCat, 2246, 0CCutri, R., et al. 2013, yCat, 2328, 0CDebes, J., Hoard, D., Kilic, M., et al. 2011, ApJ, 729, 4DDebes, J., & Sigurdsson, S. 2002, ApJ, 572, 556Debes, J., Walsh, R., & Stark, C. 2012, ApJ, 747, 148DDufour, P., Bergeron, P., Liebert, J., et al. 2007, ApJ, 663, 1291DDufour, P., Kilic, M., Fontaine, G., et al. 2010, ApJ, 719, 803DFarihi, J. 2011, in White Dwarf Atmospheres and Circumstellar Environments,

ed. D. W. Hoard (New York: Wiley), 117Farihi, J., Gansicke, B., & Koester, D. 2013, Sci, 342, 218FFarihi, J., Gansicke, B., Wyatt, M., et al. 2012, MNRAS, 424, 464Farihi, J., Zuckerman, B., & Becklin, E. 2008, ApJ, 674, 431Fontaine, G., Brassard, P., & Bergeron, P. 2001, PASP, 113, 409Hambly, N. C., Collins, R. S., Cross, N. J. G., et al. 2008, MNRAS, 384, 637Henden, A., Welch, D. L., Terrell, D., & Levine, S. E. 2009, American

Astronomical Society Meeting 214, Abstracts #407.02Hoard, D., Debes, J., Wachter, S., Leisawitz, D., & Cohen, M. 2012, ApJ,

770, 21HHolberg, J., & Bergeron, P. 2006, AJ, 132, 1221HIrwin, M. J., Lewis, J., Hodgkin, S., et al. 2004, Proc. SPIE, 5493, 411Jura, M. 2003, ApJ, 584, 91JJura, M., Dufour, P., Xu, S., et al. 2015, ApJ, 799, 109JJura, M., Farihi, J., & Zuckerman, B. 2007, ApJ, 663, 1285Jura, M., & Xu, S. 2013, AJ, 145, 30Kelson, D. 2003, PASP, 115, 688

Kelson, D., Illingworth, G., van Dokkum, P., & Franx, M. 2000, ApJ, 531, 159Kilic, M., & Redfield, S. 2007, ApJ, 660, 641KKilic, M., von Hippel, T., Leggett, S., & Winget, D. 2006, ApJ, 646, 474Kilkenny, D., O’Donoghue, D., Koen, C., Stobie, R., & Chen, A. 1997,

MNRAS, 287, 867KKoester, D. 2009, A&A, 498, 517Koester, D. 2010, MmSAIw, 81, 921Lang, D. 2014, AJ, 147, 108LLang, D., Hogg, D. W., & Schlegel, D. 2016, AJ, 151, 36LLiebert, J., Bergeron, P., & Holberg, J. 2005, ApJS, 156, 47Martin, D., Fanson, J., Schiminovich, D., et al. 2005, ApJL, 619, L1McCook, G., & Sion, E 1999, ApJS, 121, 1MMcMahon, R. G., Banerji, M., Gonzalez, E., et al. 2013, Msngr, 154, 35MO’Donoghue, D., Kilkenny, D., Koen, C., et al. 2013, MNRAS, 431, 240ORocchetto, M., Farihi, J., Gansicke, B., & Bergfors, C. 2015, MNRAS,

449, 574RSkrutski, M., Cutri, R., Stiening, R., et al. 2006, AJ, 131, 1163SVeras, D., Leinhardt, Z., Bonsor, A., & Gansicke, B. 2014, MNRAS, 445,

2244VVeras, D., Leinhardt, Z. M., Eggl, S., & Gansicke, B. T. 2015, MNRAS, 451,

3453VWilson, D., Gansicke, B., Koester, D., et al. 2015, MNRAS, 451, 3237WWright, E. L., Eisenhardt, P., Mainzer, A., et al. 2010, AJ, 140, 1868WXu, S., Jura, M., Klein, B., Koester, D., & Zuckerman, B. 2013, ApJ, 766, 132Xu, S., Jura, M., Klein, B., & Zuckerman, B. 2014, ApJ, 783, 79Zuckerman, B., Koester, D., Dufour, P., et al. 2011, ApJ, 739, 101Z

9

The Astrophysical Journal, 831:31 (9pp), 2016 November 1 Dennihy et al.